Skip to contentSkip to conversation

fs/drop_caches: move drop_caches sysctls to its own file

PATCH Superseded tangmeng 3 participants 6 messages Feb 20, 2022 – Feb 26, 2022

Subsystems (4)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

tangmeng

author
#0 lore raw

PATCH 10/11 fs/drop_caches: move drop_caches sysctls to its own file

kernel/sysctl.c is a kitchen sink where everyone leaves their dirty dishes, this makes it very difficult to maintain.

To help with this maintenance let's start by moving sysctls to places where they actually belong. The proc sysctl maintainers do not want to know what sysctl knobs you wish to add for your own piece of code, we just care about the core logic.

All filesystem syctls now get reviewed by fs folks. This commit follows the commit of fs, move the drop_caches sysctls to its own file, fs/drop_caches.c.

---
 fs/drop_caches.c   | 26 ++++++++++++++++++++++++--
 include/linux/mm.h |  6 ------
 kernel/sysctl.c    |  9 ---------
 3 files changed, 24 insertions(+), 17 deletions(-)
fs/drop_caches.c +24 -2
index e619c31b6bd9..5e044a2d75ef 100644
@@ -13,7 +13,7 @@
 #include "internal.h"
 
 /* A global variable is a bit ugly, but it keeps the code simple */
-int sysctl_drop_caches;
+static int sysctl_drop_caches;
 
 static void drop_pagecache_sb(struct super_block *sb, void *unused)
 {
@@ -47,7 +47,7 @@ static void drop_pagecache_sb(struct super_block *sb, void *unused)
 	iput(toput_inode);
 }
 
-int drop_caches_sysctl_handler(struct ctl_table *table, int write,
+static int drop_caches_sysctl_handler(struct ctl_table *table, int write,
 		void *buffer, size_t *length, loff_t *ppos)
 {
 	int ret;
@@ -75,3 +75,25 @@ int drop_caches_sysctl_handler(struct ctl_table *table, int write,
 	}
 	return 0;
 }
+
+#ifdef CONFIG_SYSCTL
+static struct ctl_table vm_drop_caches_table[] = {
+	{
+		.procname       = "drop_caches",
+		.data           = &sysctl_drop_caches,
+		.maxlen         = sizeof(int),
+		.mode           = 0200,
+		.proc_handler   = drop_caches_sysctl_handler,
+		.extra1         = SYSCTL_ONE,
+		.extra2         = SYSCTL_FOUR,
+	},
+	{ }
+};
+
+static __init int vm_drop_caches_sysctls_init(void)
+{
+	register_sysctl_init("vm", vm_drop_caches_table);
+	return 0;
+}
+late_initcall(vm_drop_caches_sysctls_init);
+#endif /* CONFIG_SYSCTL */
include/linux/mm.h +0 -6
index c3c7cb58c847..775befb2786b 100644
@@ -3126,12 +3126,6 @@ static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
 
 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
 
-#ifdef CONFIG_SYSCTL
-extern int sysctl_drop_caches;
-int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
-		loff_t *);
-#endif
-
 void drop_slab(void);
 
 #ifndef CONFIG_MMU
kernel/sysctl.c +0 -9
index b51b0b92fdc1..657b7bfe38f6 100644
@@ -2182,15 +2182,6 @@ static struct ctl_table vm_table[] = {
 		.mode		= 0644,
 		.proc_handler	= lowmem_reserve_ratio_sysctl_handler,
 	},
-	{
-		.procname	= "drop_caches",
-		.data		= &sysctl_drop_caches,
-		.maxlen		= sizeof(int),
-		.mode		= 0200,
-		.proc_handler	= drop_caches_sysctl_handler,
-		.extra1		= SYSCTL_ONE,
-		.extra2		= SYSCTL_FOUR,
-	},
 #ifdef CONFIG_COMPACTION
 	{
 		.procname	= "compact_memory",
Signature · 1 line
2.20.1
#1 lore raw

fs/drop_caches.c int drop_caches_sysctl_handler(struct ctl_table *table, int write, · line 79

Quoted 7 lines · tangmeng
fs/drop_caches.c +2 -0
@@ -75,3 +75,25 @@ int drop_caches_sysctl_handler(struct ctl_table *table, int write,
 	}
 	return 0;
 }
+
+#ifdef CONFIG_SYSCTL

fs/Makefile has:
obj-$(CONFIG_SYSCTL) += drop_caches.o

so we don't need this ifdef.

fs/drop_caches.c int drop_caches_sysctl_handler(struct ctl_table *table, int write, · line 91

Quoted 12 lines
fs/drop_caches.c +12 -0
+static struct ctl_table vm_drop_caches_table[] = {
+	{
+		.procname       = "drop_caches",
+		.data           = &sysctl_drop_caches,
+		.maxlen         = sizeof(int),
+		.mode           = 0200,
+		.proc_handler   = drop_caches_sysctl_handler,
+		.extra1         = SYSCTL_ONE,
+		.extra2         = SYSCTL_FOUR,
+	},
+	{ }
+};

Something which slightly concerns me about this sysctl splitup (which is obviously the right thing to do) is that ctl_table is quite large (64 bytes per entry) and every array is terminated with an empty one. In this example, we've gone from 64 bytes to 128 bytes.

Would we be better off having a register_sysctl_one() which registers exactly one ctl_table, rather than an array? And/or a register_sysctl_array() which takes an ARRAY_SIZE() of its argument instead of looking for the NULL terminator?

author
#2 lore raw
Quoted 12 lines · Matthew Wilcox

fs/drop_caches.c int drop_caches_sysctl_handler(struct ctl_table *table, int write, · line 79

Quoted 7 lines · tangmeng
fs/drop_caches.c +2 -0
@@ -75,3 +75,25 @@ int drop_caches_sysctl_handler(struct ctl_table *table, int write,
  	}
  	return 0;
  }
+
+#ifdef CONFIG_SYSCTL

fs/Makefile has:
obj-$(CONFIG_SYSCTL) += drop_caches.o

Quoted 1 line

so we don't need this ifdef.

Thanks for the heads up! I will delete this ifdef.

Quoted 19 lines

fs/drop_caches.c int drop_caches_sysctl_handler(struct ctl_table *table, int write, · line 91

Quoted 13 lines
fs/drop_caches.c +12 -0
+static struct ctl_table vm_drop_caches_table[] = {
+	{
+		.procname       = "drop_caches",
+		.data           = &sysctl_drop_caches,
+		.maxlen         = sizeof(int),
+		.mode           = 0200,
+		.proc_handler   = drop_caches_sysctl_handler,
+		.extra1         = SYSCTL_ONE,
+		.extra2         = SYSCTL_FOUR,
+	},
+	{ }
+};

Something which slightly concerns me about this sysctl splitup (which is obviously the right thing to do) is that ctl_table is quite large (64 bytes per entry) and every array is terminated with an empty one. In this example, we've gone from 64 bytes to 128 bytes.

Quoted 5 lines

Would we be better off having a register_sysctl_one() which registers exactly one ctl_table, rather than an array? And/or a register_sysctl_array() which takes an ARRAY_SIZE() of its argument instead of looking for the NULL terminator?

I think it is obviously the right thing that we need to do.
However, many submissions have been commited which registers an array before, I think that having a register_sysctl_one() which registers exactly one ctl_table should submit in a separate submission, rather than modify it this time.

#3 lore raw
Quoted 1 line · tangmeng

I think it is obviously the right thing that we need to do.

Since you are following up on more changes, can you work on this? Brownie points if you show size results to refelct no size difference based on a new build with an example new user.

Quoted 4 lines

However, many submissions have been commited which registers an array before, I think that having a register_sysctl_one() which registers exactly one ctl_table should submit in a separate submission, rather than modify it this time.

We can optimize this later and fix those.

Luis

author
#4 lore raw
Quoted 5 lines · Luis Chamberlain

Since you are following up on more changes, can you work on this? Brownie points if you show size results to refelct no size difference based on a new build with an example new user.

I'm going work on this, but I think it may take some times to get it done.

Before ctl_table optimization, I optimized ctl_path first. If the optimization method for ctl_path is reasonable, I will apply the optimization method to the optimization of ctl_table.

The optimization of ctl_path refers to the subject which is as follows:

[PATCH] fs/proc: Optimize arrays defined by struct ctl_path

Best regards!

Meng Tang

#5 lore raw
Quoted 8 lines · tangmeng
Quoted 5 lines · Luis Chamberlain

Since you are following up on more changes, can you work on this? Brownie points if you show size results to refelct no size difference based on a new build with an example new user.

I'm going work on this, but I think it may take some times to get it done.

Thanks and good stuff! Can you please then make that the focus of your first set of patches, and then respin this series by collecting all ACKs, Reviwed-bys? Yes please send a new v3 series for it. While at it please Cc the other folks doing sysctls changes so they are aware of this.

Thanks!

Luis